Trends in VLSI technology scaling demand that future computing devices be narrowly focused to achieve high performance and high efficiency, yet also target the high volumes and low costs of widely applicable general purpose designs. To address these conflicting requirements, we propose a modular reconfigurable architecture called Smart Memories, targeted at computing needs in the 0.1m technology generation. A Smart Memories chip is made up of many processing tiles, each containing local memory, local interconnect, and a processor core. For efficient computation under a wide class of possible applications, the memories, the wires, and the computational model can all be altered to match the applications. To show the applicability of this design, two very different machines at opposite ends of the architectural spectrum, the Imagine stream processor and the Hydra speculative multiprocessor, are mapped onto the Smart Memories computing substrate. Simulations of the mappings show that the Smart Memories architecture can successfully map these architectures with only modest performance degradation.

Smart Memories: a modular reconfigurable architecture

Operation in the subthreshold region most often is synonymous to minimum-energy operation. Yet, the penalty in performance is huge. In this paper, we explore how design in the moderate inversion region helps to recover some of that lost performance, while staying quite close to the minimum-energy point. An energy-delay modeling framework that extends over the weak, moderate, and strong inversion regions is developed. The impact of activity and design parameters such as supply voltage and transistor sizing on the energy and performance in this operational region is derived. The quantitative benefits of operating in near-threshold region are established using some simple examples. The paper shows that a 20% increase in energy from the minimum-energy point gives back ten times in performance. Based on these observations, a pass-transistor based logic family that excels in this operational region is introduced. The logic family operates most of its logic in the above-threshold mode (using low-threshold transistors), yet containing leakage to only those in subthreshold. Operation below minimum-energy point of CMOS is demonstrated. In leakage-dominated ultralow-power designs, time-multiplexing will be shown to yield not only area, but also energy reduction due to lower leakage. Finally, the paper demonstrates the use of ultralow-power design techniques in chip synthesis.

/pdf/ultralow-power-design-in-near-threshold-region-37ua2hp61u.pdf

Ultralow-Power Design in Near-Threshold Region

An electronic device (e.g., computer system, etc.) employing dynamic power management of the present invention adjusts power consumption in accordance with an analysis of parameters and events occurring over one or more time-periods. Preferably, the electronic device monitors microprocessor, operating system, peripheral and/or device-level events and adjusts run-time parameters, such as microprocessor clock frequency and voltage, to reduce power consumption with minimal perceived degradation in performance.

Method and apparatus for optimizing performance and battery life of electronic devices based on system and application parameters

Systems and methods of managing processors provide for detecting a command at a core of a processor having a plurality of cores, where the command requests a transition of the core to an idle state. Power consumption of the core is managed based on the command and an idle state status of each of the plurality of cores.

Coordinating idle state transitions in multi-core processors

A system for an integrated circuit comprising a plurality of power islands includes a first power manager and a second power manager. The first power manager manages a first power consumption for the integrated circuit based on needs and operation of the integrated circuit. The second power manager communicates with the first power manager and manages a second power consumption for one of the power islands.

Power managers for an integrated circuit

Improved operation of multi-processor chips is achieved by dynamically controlling processing load of chips and controlling, significantly greater than on/off granularity, the operating voltages of those chips so as to minimize overall power consumption. A controller in a multi-processor chip allocates tasks to the individual processors to equalize processing load among the chips, then the controller lowers the clock frequency on the chip to as low a level as possible while assuring proper operation, and finally reduces the supply voltage. Further improvement is possible by controlling the supply voltage of individual processing elements within the multi-processor chip, as well as controlling the supply voltage of other elements in the system within which the multi-processor chip operates.

Power reduction in a multiprocessor digital signal processor based on processor load

A channel decoder chip compliant with the 3GPP mobile wireless standard is described. It supports both data and voice calls simultaneously in a unified turbo/Viterbi decoder architecture. For voice services, the decoder can process over 128 voice channels encoded with rate 1/2 or 1/3, constraint length 9 convolutional codes. For data services, the turbo decoder is capable of processing any mix of rate 1/3, constraint length 4 turbo encoded data streams with an aggregate data rate of up to 2.5 Mb/s with 10 iterations per block (or 4.1 Mb/s with six iterations). The turbo decoder uses the logMAP algorithm with a programmable logsum correction table. It features an interleaver address processor that computes the 3GPP interleaver addresses for all block sizes enabling it to quickly switch context to support different data services for several users. The decoder also contains the 3GPP first channel de-interleaving function and a post-decoder bit error rate estimation unit. The chip is fabricated in a 0.18-/spl mu/m six-layer metal CMOS technology, has an active area of 9 mm/sup 2/, and has a peak clock frequency of 110.8 MHz at 1.8 V (nominal). The power consumption is 306 mW when turbo decoding a 2-Mb/s data stream with ten iterations per block and eight voice calls simultaneously.

A unified turbo/Viterbi channel decoder for 3GPP mobile wireless in 0.18-/spl mu/m CMOS

Operation of multi-processor chips is achieved by dynamically controlling processing load of chips and controlling, significantly greater than on/off granularity, the operating voltages of those chips so as to minimize overall power consumption A controller in a multi-processor chip allocates tasks to the individual processors to equalize processing load among the chips, then the controller lowers the clock frequency on the chip to as low a level as possible while assuring proper operation, and finally reduces the supply voltage Further reduction is possible by controlling the supply voltage of individual processing elements within the multi-processor chip, as well as controlling the supply voltage of other elements in the system within which the multi-processor chip operates

Power reduction in a multiprocessor digital signal processor

Power consumption in program implemented circuits and the like is dynamically controlled in accordance with the circuit performance over time, not by reducing switched capacitance as was done in prior hardware circuit implementations but, in accordance with the invention, by dynamically reducing the number of machine cycles required to implement the desired circuit at an acceptable performance level. Power consumption is dynamically reduced in program implemented circuits, for example, circuits implemented on a digital signal processor (DSP), including filters employed for channel equalization and for echo cancellation by monitoring the circuit performance over time and dynamically scaling back filter parameters when less than worst case performance is required. Scaling back of the DSP implemented filter parameters results in fewer machine cycles being required to effect the filters while maintaining adequate performance that, in turn, allows reduction in the DSP clock frequency, resulting in lower power consumption. The reduced DSP clock frequency also allows the DSP to be operated at a lower supply voltage that yields significant additional power savings.

Apparatus and method for adaptive reduction of power consumption in integrated circuits

A cache implements bank-by-bank locking to keep critical code from being flushed out of the cache. A register is maintained to rank the banks from the most recently used to the least recently used. Ordinarily, when code needs to be moved into the cache, the least recently used bank is flushed, the code is moved into that bank, and the register is updated to identify that bank as the most recently used. However, if a bank is designated in a bypass vector as being locked, that bank is bypassed in the maintenance of the register and is thus never identified as the bank to be flushed.

Christopher John Nicol

Papers

Power reduction in a multiprocessor digital signal processor based on processor load

A unified turbo/Viterbi channel decoder for 3GPP mobile wireless in 0.18-/spl mu/m CMOS

Power reduction in a multiprocessor digital signal processor

Apparatus and method for adaptive reduction of power consumption in integrated circuits

Method and memory cache for cache locking on bank-by-bank basis